The Agent Skills Directory

[PROMPT_INJECTION]: The skill processes untrusted data from GitHub issue comments via gh issue view --comments, which introduces a surface for indirect prompt injection. Malicious instructions placed in comments could potentially influence the agent's actions during the fix implementation.
- Ingestion points: External content from GitHub issue comments and descriptions retrieved in SKILL.md.
- Boundary markers: Absent; there are no instructions to the agent to distinguish between its system instructions and the untrusted data fetched from comments.
- Capability inventory: The skill allows the agent to execute sensitive operations such as terraform plan, git push, and repository-specific scripts like scripts/destroy.sh and scripts/render_harness.py.
- Sanitization: Absent; the skill does not implement filtering or validation of the fetched comment content.
[COMMAND_EXECUTION]: The skill uses various command-line utilities and scripts to perform its tasks.
- Tool usage: Standard development tools including gh, terraform, git, and ripgrep are used for project maintenance.
- Local scripts: Several project-specific scripts such as scripts/render_harness.py, scripts/contract_negative_tests.py, and scripts/destroy.sh are executed as part of the validation and cleanup process.
- AI Assistants: The skill utilizes codex and gemini CLI tools for reasoning and codebase analysis, which is appropriate for its intended use case.

fix-issue